State of Charge Estimation and Circuit Implementation for Lithium Battery Based on the Elman Neural Network Algorithm

This paper establishes an estimation model for the battery State of Charge (SOC) estimation system based on the characteristics and suitable architecture of neural network models, utilizing the Elman neural network as the central model for neural network estimation. It replaces traditional estimation methods with a neural network-based approach to identify battery state characteristics. Through the parameter training module, battery characteristic parameters are identified based on historical charging and discharging data, and the real-time estimation module is updated with these parameters to facilitate deep learning chip planning. The paper conducts system simulation parameter training and chip design planning using Dynamic Stress Tests (DST), Federal Urban Driving Schedule (FUDS), and real driving data from BMW i3 2014 BEV (SOC 90% - 10%) and BMW i3 2014 BEV (SOC 56.8% - 9.9%). The real-time estimation module of the Elman Neural Network (ENN) model is verified using SMIMS Veri Enterprise Xilinx FPGA.